[4G2-GS-7] Agents: Agents and machine learning

[4G2-GS-7-04] Modeling of Bicycle Sharing operation system with dynamic pricing by Agent Reinforcement Learning

〇Kohei Yashima¹, Setsuya Kurahashi¹ (1. University of Tsukuba)

Keywords:Bicycle sharing, Reinforcement learning, agent simulation

The purpose of this research is to encourage share cycle users to return to a specific station by paying an incentive, and to verify a business operation model that is fundamentally different from the conventional operation based on truck allocation. We employed a reinforcement learning technique for directions to user agents in order to autonomously respond to uncertain events. The value Q(s,a) used for reinforcement learning is calculated by function approximation using a hierarchical neural network. The results showed that the proposed model reduced the cost by about 20\% compared to the conventional truck operation. It suggests the usefulness of the autonomous system operation by reinforcement learning.